The LATTICE Study: Design of a pilot feasibility randomized controlled trial of lithium to delay cognitive decline in mild cognitive impairment

Abstract

INTRODUCTION

Interest has grown in lithium's neuroprotective properties in neurodegenerative illnesses. We discuss the design, rationale, and implementation of a pilot feasibility, double‐blind, randomized placebo‐controlled trial (RCT) examining whether lithium can delay cognitive decline in older adults with mild cognitive impairment (MCI).

METHODS

The study launched in September 2017. The goal was to enroll 80 community‐dwelling participants ≥ 60 years with MCI into an RCT in which they would participate for 2 years with baseline and follow‐up assessment of cognition, brain imaging, and plasma‐based biomarkers. Participants were randomized to lithium or placebo (1:1).

RESULTS

We enrolled 80 MCI older adults into the RCT. Baseline characteristics included a mean (standard deviation) age of 72 (7.7) years with 35 male and 45 female participants. Seventy‐five participants had positron emission tomography imaging for amyloid beta (Aβ), and 66 completed 7T magnetic resonance imaging. Twenty‐one participants were Aβ+ and 54 were Aβ–.

DISCUSSION

The study successfully enrolled 80 participants into an RCT examining whether lithium delays cognitive decline. The main study results will be forthcoming.

Clinical Trial Registration

NCT03185208.

Highlights

• Eighty adults ≥ 60 years with mild cognitive impairment entered a placebo‐controlled randomized controlled trial evaluating lithium's neuroprotective properties.

• Participants were followed for 2 years with baseline and follow‐up evaluations at 1 and 2 years that included neurocognitive assessment, ultra‐high‐field structural neuroimaging, positron emission tomography imaging for amyloid beta and tau, and plasma‐based biomarkers.

• Study results will be forthcoming.

1. BACKGROUND

Since US Food and Drug Administration approval in 1970, lithium has been one of the most effective treatments for bipolar disorder (BD) mood episodes. ¹ More recently, interest has grown in lithium's neuroprotective and neurotrophic properties not only in addressing cognitive impairment in BD, but also in other neurodegenerative illnesses. ² In Alzheimer's disease and Alzheimer's disease and related dementias (AD/ADRD), investigations of lithium range from animal to human to epidemiological studies, ³ including five human clinical trials, ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ three of which suggest beneficial, if modest, effects to slow cognitive decline or improve the profile of blood and cerebrospinal fluid (CSF) biomarkers. ⁴ , ⁷ , ⁹

Biological mechanisms involved in stabilizing mood may be related to lithium's anti‐dementia effects. ² Lithium inhibits glycogen synthase kinase 3 (GSK‐3; isoforms α and β), an enzyme that modulates cell survival, synaptic plasticity, cellular structure, and resilience. ¹⁰ GSK‐3α is involved in maximal processing of amyloid precursor protein to amyloid beta (Aβ); GSK‐3β phosphorylates tau. ¹¹ Additionally, through downstream effects of GSK‐3 inhibition, lithium is related to upregulation of various neurotrophic factors, including brain‐derived neurotrophic factor (BDNF) and B‐cell lymphoma protein 2 (Bcl‐2), and stabilizes dysregulated calcium homeostasis that may enhance brain health and long‐term cognitive function. ¹² Lithium treatment may be related to hippocampal neurogenesis and myelination. ¹³ Lithium has other cellular effects that include, and are not limited to, enhanced autophagy, reduced microglial activation, and reduced astrocytic inflammatory response. ¹⁴ In patients with Parkinson's disease, lithium may limit glial fibrillary acidic protein (GFAP) elevations or reduce serum neurofilament light chain (NfL) levels, ¹⁵ markers of astrocyte activation and neuroaxonal damage, respectively.

Of the human clinical trials, arguably the most comprehensive, Forlenza et al., examined lithium for amnestic mild cognitive impairment (aMCI). ⁹ They conducted a 2‐year double‐blind trial, enrolling 61 participants with aMCI, mean (standard deviation [SD]) age 72.6 (4.8) years, randomly assigning them to low‐dose lithium (target level = 0.25–0.5 mEq/L, n = 31) or placebo (n = 30). Subsequently, participants were followed single blind for an additional 24 months. Primary outcome variables were the Alzheimer's Disease Assessment Scale Cognitive subscale (ADAS‐Cog) ¹⁶ and Clinical Dementia Rating (CDR) Sum of Boxes scores. ¹⁷ Secondary outcomes included CSF concentrations of Aβ_1‐42. Fifty‐two participants completed the double‐blind trial, and 34 completed the single‐blind extension. Participants taking lithium were stable over 24 months with less deterioration in memory and attention tests, while those in the placebo group displayed cognitive and functional decline. The investigators noted that while the difference between the groups was statistically significant, it was small. Further, at 36 months, lithium treatment was associated with an increase in CSF Aβ_1‐42, suggesting to the investigators that long‐term lithium could enhance clearance of Aβ. Thirteen years after the clinical trial, the team re‐evaluated the current or last available global cognitive and functional state of a sample of the participants (11–15 years later). ¹⁸ Of the original 61, they reached 36 participants, 22 that had taken lithium and 14 who had received placebo. They found that those treated with lithium had better performance on the Mini‐Mental State Examination (MMSE) and a verbal fluency test.

Considering biological mechanisms supporting neuroprotection and previous research, we initiated LATTICE (Lithium as a Treatment to Prevent Impairment of Cognition in Elders) at the University of Pittsburgh. This study builds on and extends prior work by combining comprehensive neuropsychological assessment with plasma‐based biomarkers and advanced neuroimaging (7T magnetic resonance imaging [MRI] and positron emission tomography [PET] imaging of amyloid and tau deposits). As a pilot feasibility study, LATTICE aimed to optimize protocols for enrolling older adults with mild cognitive impairment (MCI), maintaining lithium treatment for 2 years, and completing research assessments.

In this report, we describe the design and rationale of this single‐site study examining whether lithium can delay cognitive decline in older adults with MCI. We focus on baseline data prior to randomization, discussing recruitment, interventions, participant clinical characteristics, and how we addressed challenges like the COVID‐19 pandemic. Ultimately, the study was designed to further examine whether lithium genuinely protects against dementia.

2. METHODS

We review (1) study aims and hypothesis, (2) overall design, (3) recruitment methods, (4) research procedures, and (5) the impact of the COVID‐19 pandemic on the study conduct.

2.1. Aims and hypotheses

2.1.1. Specific aims and overall study design

Our specific aim was to examine the potential disease‐modifying properties of lithium in individuals with MCI in delaying conversion to dementia.

H1: (a) Participants randomized to take lithium for 2 years, compared to placebo, will better maintain cognitive function, primarily in memory, which (b) will be associated with changes in biomarkers (e.g., GSK‐3β activity, BDNF).

H2: (a) Participants randomized to take lithium for 2 years, compared to placebo, will have larger hippocampal volumes (i.e., lower rate of reduction) and lower total gray matter thinning, which (b) will be associated with changes in biomarkers (e.g., GSK‐3β activity, BDNF) and (c) better cognitive function, primarily in memory.

Our exploratory aim was to examine whether lithium is related to additional markers of enhanced brain integrity (e.g., lower level of microbleeds, higher white matter integrity, or better network connectivity).

2.2. Overall study design

The initial design randomized 80 individuals ≥ 60 years with aMCI (single or multiple domains) to receive lithium (titrated to 0.6–0.8 mEq/L blood level if tolerated) or placebo for 2 years. This target range was selected to maximize GSK‐3 inhibition and avoid underdosing that could lead to inconclusive study results. The dosing strategy was informed by the study team's experience with a double‐blind randomized controlled trial (RCT) involving lithium in patients with BD aged ≥ 60 years. ¹⁹ In that study, participants taking lithium over 9 weeks achieved a mean blood level of 0.76 mEq/L (SD: 0.35 mEq/L). ¹⁹ Additionally, the study team was familiar with the Lithium Treatment for Agitation in Alzheimer's Disease RCT using low‐dose lithium for agitation in AD, using 300 to 600 mg daily with serum levels 0.2 to 0.6 mmol/L. ²⁰

RESEARCH IN CONTEXT

1. Systematic Review: The authors reviewed the literature using traditional (e.g., PubMed) sources and meeting abstracts and presentations. Interest has grown in lithium's neuroprotective and neurotrophic properties not only in addressing cognitive impairment in bipolar disorder, but also in other neurodegenerative illnesses.

2. Interpretation: We conducted a placebo‐controlled randomized controlled trial (RCT) in adults ≥ 60 years with mild cognitive impairment to evaluate whether lithium could delay cognitive decline and to examine its effects on brain health.

3. Future Directions: The article describes the rationale, design, and implementation of the RCT. Future publications will present the study results.

Participants underwent annual neurocognitive assessment, 7T brain MRI (high‐resolution imaging of hippocampal and cortical volumes, white matter integrity, dynamic imaging), and plasma‐based biomarker measurement. Baseline PET imaging of Aβ was conducted before randomization, with participants stratified by Aβ status using permuted blocks randomization. We projected 20% attrition, with 64 participants completing the 2‐year intervention. Figure 1 illustrates the study design and Table 1 the assessment schedule. The University of Pittsburgh Human Research Protections Office (HRPO) approved and monitored the study, which was registered with ClinicalTrials.gov (NCT03185208).

FIGURE 1.

Study design. AV‐1451, flortaucipir; Li, lithium; MRI, magnetic resonance imaging; PET, positron emission tomography; PiB, Pittsburgh compound B.

TABLE 1.

Assessments.

	Screening (T0)		RCT (T1, T2, T3)
Assessment	Initial	Comprehensive	Pre‐RCT (T1)	Quarterly	1 year (T2)	2 year (T3)
Inclusion/Exclusion	X
MINI	X
Cognitive
mTICS OR Qmci	X				X	X
3MS (In‐Person Only)	X				X	X
HSCT OR Trails A & B	X
BVMT‐R			X		X	X
CDR		X			X	X
CVLT‐II			X		X	X
DKEFS		X			X	X
E‐Cog		X			X	X
RBANS		X			X	X
Clock		X			X	X
Digit span		X			X	X
Boston Naming Test		X			X	X
NIH toolbox			X		X	X
WRAT‐4 reading		X
Clinical
BARS (monthly)				X	X	X
CIRS‐G		X			X	X
FSRP		X			X	X
Medication list		X			X	X
PASE		X		X	X	X
PHQ‐9		X		X	X	X
UKU		X		X	X	X
SRSE		X		X	X	X
PASS items	X				X	X
Imaging
7T MRI			X		X	X
PET (Aβ/tau)			X			X
Laboratory
Blood‐based Biomarkers (bi‐annual)			X		X	X
Safety labs		X		X	X	X

Abbreviations: 3MS, Modified Mini‐Mental Status; Aβ, amyloid beta; BARS, Brief Adherence Scale; BVMT, Brief Visual Memory Test‐R; CDR, Clinical Dementia Rating; CIRS‐G, Cumulative Illness Rating Scale‐Geriatric; CVLT, California Verbal Learning Test‐II; D‐KEFS, Delis–Kaplan Executive Function System; E‐Cog, Everyday Cognition scale; FSRP, Framingham Stroke Risk Profile; HSCT, Hayling Sentence Completion Test; MINI, Mini‐International Neuropsychiatric Interview; MRI, magnetic resonance imaging; mTICS, Modified Telephone Interview for Cognitive Status; NIH, National Institutes of Health; PASE, Physical Activity Scale for the Elderly; PASS, Performance Assessment of Self‐Care Skills; PET, positron emission tomography; PHQ‐9, Patient Health Questionnaire‐9 item; Qmci, Quick Mild Cognitive Impairment; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; RCT, randomized controlled trial; SRSE, spontaneous reporting of side effects; UKU, side effect rating scale, WRAT‐IV, Wide Range Achievement Test‐IV.

2.2.1. Data and safety monitoring board

As required by the National Institute on Aging (NIA), we established an external data and safety monitoring board (DSMB) with expertise in clinical trials, geriatrics, AD/ADRD, and statistics. The DSMB reviewed and approved study procedures before enrollment began (September 1, 2017) and approved all subsequent modifications. Investigators met with the DSMB semiannually. Participants’ clinical status was reviewed weekly, and all serious adverse events were reported to the HRPO within 24 hours of discovery.

2.2.2. Participant inclusion/exclusion criteria

The study had the following inclusion criteria: (1) ≥ 60 years and (2) diagnosis of MCI, according to Petersen criteria ²¹ (see section 2.4.1). Exclusion criteria included: (1) major psychiatric illness, 2) major neurologic illness, (3) contraindications to lithium, (4) inability to complete neuropsychological testing due to non‐remediable impairment (e.g., blindness). Inability to complete neuroimaging (e.g., unsafe metal implants) was initially an exclusion, but was subsequently removed in consultation with the DSMB.

2.3. Recruitment methods and screening

Prior to COVID‐19, recruitment involved presentations at senior centers, education classes, communities, and housing; internet‐based methods; partnerships with the University of Pittsburgh Alzheimer's Disease Research Center and primary care practices; and contacting former participants. During the pandemic, in‐person activities were replaced with internet and print advertisements.

All potential participants had an initial screening that was in person or by telephone. Before the COVID‐19 pandemic, our initial screening battery included three assessments: the Modified Mini‐Mental State Examination (3MS), Trail Making Test Parts A & B (TMT A/B), and the Quick Mild Cognitive Impairment (Qmci) screen. ²² , ²³ Study eligibility required participants to score beyond 1 SD below expected performance on either the Qmci, TMT A, or TMT B. Exclusion criteria included: performance exceeding 2 SDs below expected performance on two or more tests, or a 3MS score < 84 (approximately equivalent to a MMSE score of 25). ²² The Qmci was standardized for age and education variables, while TMT normative data accounted for age, education, sex, and racial factors.

The pandemic necessitated a shift to telephone‐based assessments, using the modified Telephone Interview for Cognitive Status (mTICS) ²⁴ and the Hayling Sentence Completion Test (HSCT). ²⁵ Participants qualified if they scored within the MCI range on the mTICS (total possible score: 50; > 38 indicating normal cognition, 19–38 suggesting MCI, < 19 indicating possible dementia) or achieved a scaled score of ≤ 4 on any HSCT component (Section 1, Section 2, Section 2 error count, or overall score). These telephone‐based measures were administered without demographic normative adjustments.

If the potential participant was grossly eligible (e.g., scoring in the MCI range on cognitive tests and having no excluding conditions), they were further screened in person with the Mini‐International Neuropsychiatric Interview (MINI) ²⁶ and the medication management portion of the Performance Assessment of Self‐Care Skills (PASS) to assess ability to safely manage the study medication. ²⁷ Individuals who passed the initial screen signed consent, according to the Declaration of Helsinki, and underwent more comprehensive assessment as described below.

2.4. Research procedures

2.4.1. Neurocognitive assessment

Individuals meeting preliminary MCI criteria through screening, described above, subsequently underwent comprehensive neuropsychological evaluation. The evaluation involved the CDR scale, Everyday Cognition Scale (Ecog) with both self‐ and informant reports, the Wide Range Achievement Test 4th edition (WRAT‐4) Reading subtest, Boston Naming Test, Clock Drawing Test, the Wechsler Adult Intelligence Scale (WAIS) IV Digit Span subtest, the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS; using two counterbalanced alternative forms to minimize practice effects), and select Delis–Kaplan Executive Function System (D‐KEFS) subtests (Verbal Fluency, TMT, and Color Word Interference). ²² , ²⁸ Functional assessment incorporated the PASS, ²⁷ focusing on shopping, medication management, and bill payment tasks—domains previously identified as sensitive in distinguishing between normal cognition and mild cognitive impairment. ²⁹ Trained psychometrists administered the assessment battery over 4 to 5 hours with breaks to prevent fatigue.

Cognitive diagnostic determination involved a multidisciplinary adjudication conference including a neuropsychologist, neurologist, and geriatric psychiatrist. The team applied the National Alzheimer's Coordinating Center/Revised Petersen comprehensive criteria for MCI diagnosis. ²¹ Cognitive test performances were standardized using demographic‐adjusted normative data (accounting for age, education, race, and/or sex where available) and interpreted in the context of estimated literacy (WRAT‐4 Reading), education background, occupation history, and relevant medical information. MCI diagnosis required: (1) subjective cognitive concern reported by participant or informant on the Ecog, (2) objective cognitive impairment evidenced by performance 1 to 2 SDs below expectation on either two tests within a single cognitive domain or three tests across multiple cognitive domains, (3) relatively preserved functional independence as demonstrated by CDR and PASS performance, and (4) absence of dementia‐level cognitive impairment. In cases with minimal subjective cognitive complaints but sufficient objective evidence satisfying criteria 2 to 4, an MCI diagnosis was still assigned based on education and occupation considerations.

Prior to entering the RCT, participants received the Brief Visual Memory Test Revised (BVMT‐R), California Verbal Learning Test 2nd edition (CVLT‐II), and National Institutes of Health (NIH) Toolbox Cognition subtests. ²² , ³⁰ Except for the WRAT‐4 Reading, all cognitive tests were completed during the comprehensive screening at T1 and were repeated at 1 and 2 years. The main cognitive outcome, a measure of overall cognitive function, included the 3MS (global cognitive screen), RBANS delayed list recall (verbal memory) and coding (processing speed), D‐KEFS TMT condition 4 (set shifting/executive function), and performance on cognitive‐instrumental activities of daily living (IADL) tasks measured with the PASS (high level everyday function). These tests were organized into a composite variable, modeled after the Preclinical Alzheimer Cognitive Composite (PACC) ³¹ (see section 2.4.7). The pre‐specified cognitive outcomes for hypothesis testing included overall cognitive function (PACC), verbal memory (CVLT‐II), and visual memory (BVMT‐R).

2.4.2. MR imaging

The study team used ultra‐high‐field (7T) MRI and a customized radiofrequency system to detect brain pathologies in AD. ³² Compared to 3T MRI, 7T has greater sensitivity and spatial resolution for detecting changes in cortical and white matter structure and integrity. ³³ For structural outcome measures, the study has focused on hippocampal volumes and total cerebral cortical gray matter (primary outcomes). Additional markers of brain integrity included other AD regions (amygdala, entorhinal cortex, and temporal lobe volumes), cerebral white matter, and white matter hyperintensity (WMH) burden. Baseline volumetric analysis was conducted using T1‐weighted images acquired with 0.75 mm isotropic resolution and processed using the FreeSurfer package (version 7.1.1) pipeline adapted for 7T imaging. ³³ WMH burden was quantified using the wmh_seg package (https://github.com/jinghangli98/wmh_seg) and T2‐weighted fluid‐attenuated inversion recovery images. For all MRI‐derived volumetric data, visual quality assurance was performed, and manual corrections to the segmentations were made when necessary. An overview of the 7T MRI methods for brain morphometrics and WMH segmentation is shown in Figure S1 in supporting information.

2.4.3. PET imaging

PET/MR imaging was performed on a Siemens Biograph mMR scanner (Siemens Medical Systems USA) with simultaneous 3T MR capability. ³⁴ Cerebral Aβ burden was assessed using [11C] Pittsburgh compound B (PiB) PET (15.0 mCi IV dose) with 50 to 70 minutes post‐injection acquisition. ³⁵ Concurrent MR included T1‐weighted magnetization‐prepared rapid gradient echo (MPRAGE) and Dixon sequences. [11C]PiB PET and MPRAGE analysis used a FreeSurfer‐based pipeline yielding standardized uptake value ratios (SUVRs) for a composite nine‐region measure (GBL9). ³⁶ Aβ positivity was defined as GBL9 SUVR ≥ 1.346, a threshold established via sparse k‐means clustering and resampling ³⁷ from an independent cohort of 61 cognitively normal participants. Tau imaging used [18F]AV‐1451 (10 mCi) acquired 80 to 100 minutes post‐injection, ³⁶ registered to 3T MPRAGE images, and analyzed with identical FreeSurfer parcellation. Tau positivity was defined as MetaTemporal SUVR > 1.18, ³⁸ a region optimal for detecting early AD‐specific pathology. ³⁹

2.4.4. Blood samples for safety monitoring and biomarker assessment

Safety labs included a basic metabolic panel (BMTP), thyroid stimulating hormone (TSH), and EKG at baseline, 1‐year follow‐up (T2), and 2‐year follow‐up (T3); urinalysis and urine osmolality at baseline and T3; lithium levels at biweekly titration visits; BMTP, TSH, and lithium levels at quarterly visits and additionally if clinically indicated. Data and safety monitoring included weekly review of all participants’ safety labs to identify any adverse effects on renal, thyroid, or parathyroid function. For biomarker measures, we obtained blood at study entry, then every 6 months for analysis of apolipoprotein E (APOE) genotype (baseline only), GSK‐3 activity, BDNF, and other exploratory AD/ADRD plasma biomarkers (e.g., Aβ42, Aβ40, GFAP, NfL, and phosphorylated tau [p‐tau]217). The NULISAseq CNS Disease Panel 120 assay (https://alamarbio.com/products‐and‐services/nulisaseq‐cns‐disease‐panel/) was conducted at the Biofluid Biomarker Laboratory, Department of Psychiatry, School of Medicine, University of Pittsburgh, Pittsburgh, PA, USA, using validated kits (Ref # 800104) on an Alamar ARGO system following published protocols. ⁴⁰

2.4.5. Additional assessments

The study assessed medical comorbidity (Cumulative Illness Rating Scale‐Geriatrics [CIRS‐G]) ⁴¹ and specifically cardiovascular risk factors (Framingham Stroke Risk Profile [FSRP]), ⁴² mood (Patient Health Questionnaire 9 item [PHQ‐9]), ⁴³ physical activity (Physical Activity Scale for the Elderly [PASE]), ⁴⁴ study drug compliance (Brief Adherence Rating Scale [BARS]), ⁴⁵ and side effects (UKU side effect rating scale ⁴⁶ and spontaneous reporting of side effects [SRSE]).

Anticholinergic burden of medications was calculated for each participant based on their medication list as previously described. ⁴⁷ Briefly, medications were rated on anticholinergic activity as follows: score 0 (medications with no known anticholinergic activity), score 1 (drugs with serum anticholinergic activity but without negative cognitive effects), score 2 (drugs with clinically relevant anticholinergic activity), and score 3 (drugs with high anticholinergic potency). Anticholinergic burden was the sum of anticholinergic activity of all the medications a participant was taking.

2.4.6. Study medications, blinding, and laboratory monitoring

Following procedures developed in the Acute Pharmacotherapy of Late‐Life Mania study, ¹⁹ the University of Pittsburgh Investigational Drug Service (IDS) acquired lithium in 150 and 300 mg capsules and over‐encapsulated them to mask the contents. The IDS also created matching placebos. The study team instructed participants to start the study medication, lithium 150 mg or placebo, at one pill daily or every other day, based on general medical status and concomitant medications. The team adjusted dosing on a weekly basis in conjunction with evaluation of tolerability and side effects. The team assessed side effects with open‐ended queries of the participants supplemented by the UKU and adherence with the BARS.

To preserve medication masking, all participants underwent lithium blood level monitoring regardless of group assignment. An unblinded team member, who had no other interaction with the team or participants except during medical emergencies requiring unblinding, obtained the results and reported either real or algorithm‐generated false levels to the team.

2.4.7. Study modifications

With DSMB approval, we made three key modifications during implementation. First, in April 2018, facing recruitment challenges, we expanded inclusion to individuals with non‐amnestic MCI with attention/executive dysfunction. We reasoned that some non‐amnestic MCI participants might be Aβ+ and, because randomization was already stratified by Aβ status with pre‐specified Aβ+ subgroup analyses, this modification was methodologically sound. Additionally, lithium's pleiotropic effects might benefit those with vascular cognitive impairment. ⁴⁸ Second, we modified our assessment of overall cognition. Instead of relying solely on RBANS and D‐KEFS subtests, we implemented a PACC score that incorporated these tests along with a global cognitive screen and cognitive‐IADLs assessment. This change enhanced our measurement of cognitive changes and aligned our methods with other trials examining cognition in older adults. Third, with DSMB encouragement, we prioritized medication tolerability rather than achieving target lithium blood levels (0.6–0.8 mEq/L) to maximize retention. Most participants could not reach or maintain these target levels due to side effects (including tremor, hair loss, and stomach upset) or concerns about worsening kidney or impaired thyroid function. As a result, some participants required doses as low as 75 mg daily or less. Other modifications included: eliminating CSF evaluation due to participant refusal; allowing entry without neuroimaging for otherwise eligible participants; including those on cholinesterase inhibitors or memantine if adjudicated as MCI; designating NIH Toolbox as exploratory as it lacks composite scores for those > 85; and implementing pandemic safety protocols. Figure 2 shows a timeline of key milestones.

FIGURE 2.

Timeline of key milestones. DSMB, data safety monitoring board; MCI, mild cognitive impairment; PACC, Preclinical Alzheimer Cognitive Composite; RCT, randomized controlled trial.

2.4.8. Statistical methods and power

Analysis will begin with descriptive methods and two‐sample testing (t tests, Wilcoxon, chi ‐squared) to compare intervention and control groups. Two‐sided tests will use a 0.05 significance level. Primary longitudinal analyses will use linear mixed effects models including time, group, and time‐by‐group interaction terms.

For Hypotheses 1a/2a, power calculations based on 80 randomized subjects (64 completers, 32 per group) with three measurement timepoints indicate that using a two‐sided significance level of 0.05 and 0.80 power, we can detect a medium effect size (Cohen d = 0.57) when observations have a 0.5 correlation across time.

2.4.9. Data sharing

Results will be published and posted on www.clinicaltrials.gov. Data will be shared with NIA and qualified investigators through National Centralized Repository for Alzheimer's Disease (https://ncrad.iu.edu), with request instructions posted on ClinicalTrials.gov within 12 months of study completion.

2.4.10. COVID‐19

On March 20, 2020, the University of Pittsburgh restricted all on‐campus research work to essential activities due to the COVID‐19 pandemic. All recruitment and in‐person activities stopped. The University of Pittsburgh allowed limited research procedures on enrolled participants to resume on July 21, 2020, and, with appropriate precautions, full research procedures on December 21, 2020. While the study was originally planned to end May 31, 2022, the COVID‐19 pandemic resulted in disruptions that extended enrollment and delayed study completion until 2024. We created assessment forms to track all SARS‐CoV‐2 infections and symptoms, including long‐term symptoms. Four participants entered the study, meeting all inclusion/exclusion criteria, after having COVID‐19.

3. RESULTS

Despite COVID‐19 disruptions, we enrolled 80 participants into the RCT. From DSMB approval (September 1, 2017) through the last participant enrollment (August 29, 2022), we received 838 referrals and screened 522 individuals. Of these, 352 were excluded (Table 2), and 170 signed informed consent (demographics in Table 3A). After consent, 90 participants withdrew, were excluded, or were lost to follow‐up (Figure 3), including 3 participants who were randomized but did not start study medication. Tables 3A, 3B, 3C, and 3D. present demographic and clinical characteristics of the 80 RCT participants. We present clinical and imaging results stratified by aMCI and non‐amnestic MCI in Table S1 in supporting information. Demographics were similar between all consented participants and those randomized to the RCT, with groups A and B being well balanced. One participant in Group B did not receive the BVMT‐R and CVLT‐II because she had substantial familiarity with those tests. Three Group A participants had missing data, preventing calculation of their PACC scores. Four participants had COVID‐19 before study entry (one in Group A, three in Group B). Neuroimaging completion rates were 81% for 7T MRI (n = 65), 89% for amyloid PET (n = 71), and 45% for tau PET (n = 36). See Table S2 in supporting information for details. Last, we could not evaluate GSK‐3β as commercial assays failed to reliably measure its activity in plasma. Future studies will require custom‐developed plasma assays.

TABLE 2.

Participants excluded prior to signing consent.

N	Reasons for screen failure
105	Normal cognition
66	Withdrew
48	Medical contraindication (separate from impaired kidney function)
44	Non‐compliant (e.g., canceled scheduled appointments)
39	Dementia
17	Psychiatric diagnosis
14	Unreachable
10	Impaired kidney function
5	Visual or auditory impairment
3	Lived too far away to participate
1	Unable to complete testing because English was second language
352	Total

TABLE 3A.

Demographic characteristics of individuals who signed consent, participants who were randomized and entered the trial, and participants in the trial based on masked group assignment (A or B).

	All enrolled (signed consent)	Randomized and started medication	Group A	Group B
N (%)	170 (100)	80 (100)	39 (48.8)	41 (51.2)
Age	71.37 (7.80)	72.10 (7.73)	71.22 (6.47)	72.93 (8.77)
Sex N (%)
Male	74 (43.5)	35 (43.8)	17 (43.6)	18 (43.9)
Female	96 (56.5)	45 (56.3)	22 (56.4)	23 (56.1)
Education	15.83 (2.33)	15.99 (2.33)	16.54 (1.80)	15.46 (2.66)
Race N (%)
White	150 (88.2)	71 (88.8)	33 (84.6)	38 (92.7)
Black/African American	18 (10.6)	8 (10.0)	5 (12.8)	3 (7.3)
Asian	2 (1.2)	1 (1.2)	1 (2.6)	0 (0)
Ethnicity N (%)
Hispanic or Latino	1 (0.6)	1 (1.3)	1 (2.6)	0 (0)
Not Hispanic or Latino	169 (99.4)	79 (98.8)	38 (97.4)	41 (100.0)

FIGURE 3.

Outcome of 170 individuals who signed informed consent. MCI, mild cognitive impairment; RCT, randomized controlled trial.

TABLE 3B.

Clinical characteristics.

	Started med	Group A	Group B
Aβ–	54	27	27
Aβ+	21	10	11
Aβ unknown	5	2	3
MCI N (%)
Amnestic	62 (77.5)	28 (71.8)	34 (82.9)
Not amnestic	18 (22.5)	11 (28.2)	7 (17.1)
APOE ε4 N (%)
Non‐carriers	52 (65.0)	26 (66.7)	26 (63.4)
Carriers	28 (35.0)	13 (33.3)	15 (36.6)
	Mean (SD)
FSRP	0.12 (0.12)	0.11 (0.12)	0.13 (0.12)
CIRS‐G total	10.50 (4.20)	9.90 (3.61)	11.07 (4.67)
Creatinine (mg/dL)	0.88 (0.15)	0.92 (0.14)	0.84 (0.16)
GFR (mL/min/1.73m2)	81.17 (12.74)	77.95 (13.11)	84.23 (11.74)
Medication count	7.23 (5.04)	6.64 (4.84)	7.78 (5.22)
Anticholinergic burden	2.44 (2.61)	2.34 (2.85)	2.53 (2.39)
PASE	103.77 (64.47)	90.20 (50.91)	117.01 (73.65)
PHQ‐9	3.69 (3.35)	3.49 (3.46)	3.88 (3.27)

Abbreviations: Aβ, amyloid beta; APOE, apolipoprotein E; CIRS‐G, Cumulative Illness Rating Scale‐Geriatric; FSRP, Framingham Stroke Risk Profile; GFR, glomerular filtration rate; MCI, mild cognitive impairment; PASE, Physical Activity Scale for the Elderly; PHQ, Patient Health Question 9 items; SD, standard deviation.

TABLE 3C.

Neuropsychological and neuroimaging characteristics.

Mean (SD) (N, if reduced)	Started med	Group A	Group B
BVMT‐R a	6.39 (2.98) [79]	6.56 (2.86) [39]	6.23 (3.11) [40]
CVLT‐II a	7.92 (3.63) [79]	7.90 (3.90) [39]	7.95 (3.40) [40]
PACC score a	0.00 (3.39) [77]	0.41 (3.16) [36]	−0.36 (3.59) [41]
Tau SUVR	1.11 (0.14) [34]	1.12 (0.14) [13]	1.11 (0.13) [21]
Brain volume (mm3)	[62]	[29]	[33]
Hippocampal a	7299.37 (1086.53)	7199.39 (801.07)	7387.23 (1293.01)
Amygdala	2996.11 (530.06)	3029.30 (455.27)	2966.95 (593.60)
Entorhinal	2545.27 (566.82)	2459.03 (490.31)	2621.06 (624.05)
Temporal lobe	71721.47 (9016.24)	70764.28 (7624.44)	72562.64 (10126.13)
Cerebral cortical gray matter a	414052.64 (46467.75)	410702.08 (40439.38)	416997.07 (51633.65)
Cerebral white matter	406103.21 (62226.10)	404877.07 (60644.66)	407180.72 (64502.35)
WMH (mm3)	7201.57 (10107.22) [63]	6773.88 (6271.82) [30]	7590.37 (12726.02) [33]

Abbreviations: Aβ, amyloid beta; BVMT‐R, Brief Visual Memory Test revised; CVLT, California Verbal Learning Test‐II (long delayed free recall); BVMT, Benton Visual Memory Test‐revised (delayed recall); PACC, Preclinical Alzheimer Cognitive Composite; SD, standard deviation; SUVR, standardized uptake value ratio; WMH, white matter hyperintensity.

^a Primary outcome.

TABLE 3D.

Plasma biomarker data.

	Started med	Group A	Group B
	N = 73	N = 37	N = 36
Mean (SD) (values are log transformed)
Aβ40	11.90 (0.70)	12.01 (0.66)	11.79 (0.73)
Aβ42	13.07 (0.94)	13.03 (0.98)	13.11 (0.90)
Aβ42/Aβ40	1.10 (0.09)	1.09 (0.09)	1.11 (0.08)
BDNF a	14.51 (1.19)	14.41 (1.47)	14.61 (0.83)
GFAP	13.11 (0.68)	13.17 (0.65)	13.05 (0.71)
NfL	11.75 (0.66)	11.80 (0.74)	11.70 (0.57)
p‐tau217	11.70 (0.52)	11.72 (0.55)	11.68 (0.49)

Abbreviations: Aβ, amyloid beta; BDNF, brain‐derived neurotrophic factor; GFAP, glial fibrillary acidic protein; NfL, neurofilament light chain; p‐tau, phosphorylated tau; SD standard deviation.

^a Primary outcome.

Most participants in the RCT presented with memory impairments. Of the total cohort, 62 participants (77.5%) had aMCI (either single or multiple domains), while 18 participants (22.5%) had non‐amnestic MCI. Regarding Aβ status, 21 participants (26%) were Aβ+, 54 (68%) were Aβ–, and 5 (6%) did not undergo amyloid imaging. All Aβ+ participants had amnestic disturbance. Among the 62 participants with amnestic disturbance, 59 completed amyloid PET imaging, with 21 (34%) classified as Aβ+ and 38 (61%) as Aβ–.

4. DISCUSSION

In this report, we have presented the design, rationale, and challenges of implementing a pilot feasibility study of lithium for delaying cognitive decline in older adults with MCI. Additionally, we have reviewed the impact of the COVID‐19 pandemic on the study conduct. Due to the COVID‐19 pandemic, for months we were unable to enroll participants and perform in‐person procedures. Although there were a few missed assessments overall, many participants experienced delays in their scheduled evaluations, and some in‐person assessments had to be converted to telephone format. Despite the pandemic‐related disruptions, the study team enrolled 80 participants into the RCT.

Since the study was launched in 2017, the state of the science has rapidly advanced. For example, were the study to be initiated now, instead of using PET imaging for Aβ for screening or RCT stratification, we would use plasma‐based biomarker assessment for screening and eligibility. Plasma‐based biomarker tests, such as p‐tau217, could be used to rule out potential participants, after a brief initial cognitive screening, who would be very unlikely to have biological evidence of AD. ⁴⁹ This would be cost and time efficient, if enrolling participants with pre‐clinical or prodromal AD. ³⁹ We note that the proportion of Aβ+ participants was low (26%), likely due to recruitment outside of a memory disorders clinics where prodromal AD is less prevalent, and MCI may have non‐AD etiology. ⁵⁰

The main study results will be forthcoming. If we identify that lithium preserves memory, overall cognition, or brain integrity, then this would warrant a larger confirmatory study. If there is no effect or if participants worsen with lithium, then this might suggest lithium should not be pursued further or that there might have been methodological flaws, such as insufficient therapeutic lithium concentration, low rate of amyloid positivity and inclusion of participants with non‐AD pathology, unbalanced distribution of confounders (e.g., COVID‐19 accelerating cognitive decline or AD pathology), or other limitations. Ideally, the results of the LATTICE study, by integrating neurocognitive assessment, advanced neuroimaging, and plasma biomarkers, will enhance our understanding of whether lithium has neuroprotective properties.

CONFLICT OF INTEREST STATEMENT

In the past 36 months, Dr. Gildengers received an honorarium for an invited lecture (Inaugural Ashok Gajwani, M.D., Lectureship at the 2024 UTHealth Houston Mood Disorders Conference). Dr. Karikari has received honoraria for invited lectures (University of Wisconsin‐Madison, University of Pennsylvania, NIH, CQDM Canada). He was a consultant to Quanterix, Neurogen Biomarking, and SpearBio. He holds a patent for the use of a pS39g assay to diagnose tauopathies. Dr. Lopez was a consultant for Novo Nordisk. All other authors report no conflicts of interest in connection with this report. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All human subjects provided informed consent.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Gildengers AG, Ibrahim TS, Zeng X, et al. The LATTICE Study: Design of a pilot feasibility randomized controlled trial of lithium to delay cognitive decline in mild cognitive impairment. Alzheimer's Dement. 2025;11:e70112. 10.1002/trc2.70112